Introduction

Appropriate antibiotics may improve survival in patients with bacterial pneumonia. However, the use of empirical broad-spectrum antibiotics in patients without infection is potentially harmful, facilitating colonization and superinfection with multiresistant micro-organisms [1]. In order to achieve these goals most authors recommend starting broad-spectrum antibiotics immediately after a bacteriological sample (plus one or two blood cultures) has been performed [1–5]. This strategy permits a rapid stop of the antibiotic in the event of negative findings and a subsequent de-escalation according to the micro-organisms recovered from bacteriological culture [2, 5, 6]. Unfortunately, bacteriology is not available at any time in most ICUs. Consequently one major question is the reliability of bacteriological samples performed after the institution of new antibiotics in patients with recent pulmonary infiltrates compatible with nosocomial pneumonia. The impact of previous antimicrobials used on this strategy is a key matter of concern.

Current antibiotic therapy

It is not infrequent that patients being treated for extrapulmonary infection develop findings compatible with ventilator-associated pneumonia (VAP). If not modified within the previous 72 h, antimicrobials are then defined as current antimicrobials. When nosocomial pneumonia occurs in such patients, the causal organisms are likely to be resistant to these antibiotics and their growth unaffected by them. This finding has been supported by two studies with the judgment of clinical experts as the standard [7, 8]. It is also supported by the only postmortem study, which clearly states that antimicrobials were unchanged in the past 3 days [9].

Recent antibiotic therapy

On the other hand, when antimicrobials have been started within the previous 24 h, they are defined as recent antimicrobials [7]. In such cases, as compared to the judgment of clinical experts, the sensitivity of protected specimen brush (PSB) and bronchoalveolar lavage culture (BAL-C) dramatically decreases. It is only 38% with BAL-C and 40% with PSB. Similarly, several authors have investigated follow-up bacteriological sampling after the beginning of antimicrobials instituted to treat VAP. Prats et al. [10] performed serial PSB 12, 24, 48, and 72 h after appropriate antibiotic therapy was begun. They found that the mean number of colony-forming units per ml (cfu/ml) falls from 80,300 ± 35,200 to 15,500 ± 27,130 after only 12 h of antibiotics. Prats et al. suggested that sensitivity can be maintained at an acceptable level by decreasing the threshold of the PSB. Similar results have been showed for endotracheal aspirates (EA). Dennesen et al. [11] found a decrease of more than 2 log 10 cfu/ml in the culture after 3 days.

In practice, when antimicrobials have been recently modified, a negative finding indicates either that the patient has been successfully treated for pneumonia and the bacteria are eradicated (but de-escalation is impossible), or that the lung infection was not present to begin with (leading to an active research of other diseases and cessation or modification of antibiotics).

Areas of uncertainties

However, areas of uncertainties persist. Firstly, the gold standard used to diagnose VAP in the above studies was clinical judgment or other microbiological investigations. No report currents exists which used histology and bacteriology of the lung as a standard. Secondly, the impact of recent antimicrobials may differ according to the type of bacteriological samples. BAL could possibly be less affected than other quantitative techniques. Rouby et al. [12] compared 30 patients with histologically confirmed VAP and 29 free of pneumonia. All patients with suspected VAP received antimicrobials before they died. The sensitivity of protected nonbronchoscopic BAL (P-BAL) vs. postmortem lung histology and culture was 80% and specificity 66%. In 73% of cases the micro-organisms were partially or completely in accordance with those recovered from the lung biopsy specimen. Similarly, Torres et al. [13] compared guided lung biopsy histology and cultures (≥ 103 cfu/g) to conventional BAL, protected BAL, PSB, and EA in 25 patients. The previous duration of antibiotic therapy (17/25 patients) was 9.5 ± 7.9 days, and the purposes of previous antimicrobials were not detailed. Previous antimicrobials decrease both sensitivity and specificity of EA (≥ 105 cfu/ml: on AB sensitivity 50%, specificity: 71%, off AB sensitivity 75%, specificity 100%), and of PSB (≥ 103 cfu/ml: on AB sensitivity 50%, specificity: 86%, off AB sensitivity 87%, specificity 100%). On the other hand, the accuracy of conventional BAL (3104 cfu/ml: on AB sensitivity 83%, specificity 75%, off AB sensitivity 62%, specificity 80%) and of protected BAL (≥ 104 cfu/ml on AB sensitivity 67%, specificity: 89%, off AB sensitivity 25%, specificity 100%) remained unchanged [12, 13]. Similarly, in the study from Souweine et al. [7] the accuracy of the percentage of intracellular organisms in BAL fluid (ICO count: threshold ≥ 5%) was not affected by previous antimicrobials (no AB 71%, current AB 50%, recent AB 67%).

Thirdly, the impact of antimicrobials on the accuracy of bacteriological samplings depends on the minimal inhibitory concentrations of antimicrobials used and the pharmacokinetics of the antimicrobials. The bronchial concentration of molecules such as vancomycin is low and may explain partly why protected BAL remains positive after 3–5 days of vancomycin in methicillin-resistant Staphylococcus aureus VAP [14]. Similarly, aminoglycosides poorly penetrates in the lung. It should therefore little affect the accuracy of samplings. Conversely, antimicrobial such as fluoroquinolones may concentrate in the bronchial mucosa and inhibit the growth of some, even nonsusceptible, bacteria [15]. Fourthly, antimicrobials such as aminoglycosides or fluoroquinolones or imipenem (on Pseudomonas aeruginosa) have an important postantibiotic effect which might influence the accuracy of the examinations for a longer period than other antimicrobials. Finally, the impact is more important for the most susceptible bacterias. Previous studies found that the decrease in the bacterial count were more pronounced for Streptococcus pneumoniae  [10, 11], Haemophilus influenzae [11], or for methicillin-sensitive S. aureus than for methicillin-resistant S. aureus [14], P. aeruginosa [10, 11], or Acinetobacter baumannii [10]. Discordant results were obtained for Enterobacteriaceae [10, 11].

In Intensive Care Medicine da Silva and coworkers [16] now investigate the diagnostic value of the direct examination with intracellular organism (ICO) count and the culture of bronchoalveolar lavage in an animal model of bacterial pneumonia due to either S. pneumoniae or P. aeruginosa. They evaluate the impact of a 48-h single-antibiotic treatment on the accuracy of the technique and highlight the impact of recent antimicrobials on BAL results. Using histology as the gold standard, they found that the sensitivity of BAL-C is largely decreased by treatment with penicillin in the case of pneumococcal pneumonia. To a lesser extent, similar results were obtained after a pretreatment of P. aeruginosa pneumonia by ceftazidime. On the other hand, the effect of a single therapy by amikacin was negligible on the sensitivity of BAL-C in the same model. Finally they confirmed that a ICO count of less than 2% retains a large part of its accuracy in diagnosing pneumonia in such cases.

However, the results of the experiments performed in a rat model need to be interpreted with caution. For example, the P. aeruginosa experiments leads to a less important histological pneumonia and a relatively poor accuracy of BAL culture, even when no treatment were given, which is surprising and needs further research. Moreover, the minimal inhibitory concentrations of the strains used and the level of antimicrobials obtained in the lung parenchyma are not given.

What could we propose in the case of VAP suspicion and previous antibiotic therapy?

To conclude, many techniques, either blind or directed by a bronchoscopy, are available for the management of nosocomial pneumonia, and must be performed before the initiation of any new antimicrobial treatment for suspected pneumonia (Fig. 1). Even after a recent antibiotic treatment the BAL culture and especially the ICO count retain part of their diagnostic properties. Except for the most susceptible ones the bacteria will be recovered in the BAL culture but frequently in concentrations under the classical threshold of 104 cfu/ml. However, de-escalation of antibiotic treatment is difficult in this cases [2]. To cope with this problem, if bacteriological analyses are not available in emergency, it should be known that a bacteriological specimen could be processed after refrigeration with a good reliability [17–19]. After 24 and even 48 h refrigeration at +4 °C BAL culture remains an acceptable alternative as it specificity is unchanged and its sensitivity falls only slightly [18, 19].

Fig. 1
figure 1

Proposed diagnostic strategy in the case of previous antibiotic therapy. BAL, Bronchoalveolar lavage; ICO, intracellular organisms; Abx, antibiotics

Full size image